1 .\" Copyright (C) 2019 Michael Kerrisk <mtk.manpages@gmail.com>
2 .\" A very few fragments remain from an earlier page written by
3 .\" Werner Almesberger in 2000
5 .\" %%%LICENSE_START(VERBATIM)
6 .\" Permission is granted to make and distribute verbatim copies of this
7 .\" manual provided the copyright notice and this permission notice are
8 .\" preserved on all copies.
10 .\" Permission is granted to copy and distribute modified versions of this
11 .\" manual under the conditions for verbatim copying, provided that the
12 .\" entire resulting derived work is distributed under the terms of a
13 .\" permission notice identical to this one.
15 .\" Since the Linux kernel and libraries are constantly changing, this
16 .\" manual page may be incorrect or out-of-date. The author(s) assume no
17 .\" responsibility for errors or omissions, or for damages resulting from
18 .\" the use of the information contained herein. The author(s) may not
19 .\" have taken the same level of care in the production of this manual,
20 .\" which is licensed free of charge, as they might when working
23 .\" Formatted or processed versions of this manual, if unaccompanied by
24 .\" the source, must acknowledge the copyright and authors of this work.
27 .TH PIVOT_ROOT 2 2021-03-22 "Linux" "Linux Programmer's Manual"
29 pivot_root \- change the root mount
32 .BR "#include <sys/syscall.h>" " /* Definition of " SYS_* " constants */"
33 .B #include <unistd.h>
35 .BI "int syscall(SYS_pivot_root, const char *" new_root \
36 ", const char *" put_old );
40 glibc provides no wrapper for
42 necessitating the use of
46 changes the root mount in the mount namespace of the calling process.
47 More precisely, it moves the root mount to the
48 directory \fIput_old\fP and makes \fInew_root\fP the new root mount.
49 The calling process must have the
51 capability in the user namespace that owns the caller's mount namespace.
54 changes the root directory and the current working directory
55 of each process or thread in the same mount namespace to
57 if they point to the old root directory.
61 does not change the caller's current working directory
62 (unless it is on the old root directory),
63 and thus it should be followed by a
64 \fBchdir("/")\fP call.
66 The following restrictions apply:
76 must not be on the same mount as the current root.
78 \fIput_old\fP must be at or underneath \fInew_root\fP;
79 that is, adding some nonnegative
80 number of "\fI/..\fP" prefixes to the pathname pointed to by
82 must yield the same directory as \fInew_root\fP.
85 must be a path to a mount point, but can't be
87 A path that is not already a mount point can be converted into one by
88 bind mounting the path onto itself.
90 The propagation type of the parent mount of
92 and the parent mount of the current root directory must not be
96 is an existing mount point, its propagation type must not be
98 These restrictions ensure that
100 never propagates any changes to another mount namespace.
102 The current root directory must be a mount point.
104 On success, zero is returned.
105 On error, \-1 is returned, and
106 \fIerrno\fP is set to indicate the error.
109 may fail with any of the same errors as
111 Additionally, it may fail with the following errors:
114 .\" Reconfirmed that the following error occurs on Linux 5.0 by
115 .\" specifying 'new_root' as "/rootfs" and 'put_old' as
116 .\" "/rootfs/oldrootfs", and *not* bind mounting "/rootfs" on top of
117 .\" itself. Of course, this is an odd situation, since a later check
118 .\" in the kernel code will in any case yield EINVAL if 'new_root' is
119 .\" not a mount point. However, when the system call was first added,
120 .\" 'new_root' was not required to be a mount point. So, this
121 .\" error is nowadays probably just the result of crufty accumulation.
122 .\" This error can also occur if we bind mount "/" on top of itself
123 .\" and try to specify "/" as the 'new' (again, an odd situation). So,
124 .\" the EBUSY check in the kernel does still seem necessary to prevent
125 .\" that case. Furthermore, the "or put_old" piece is probably
126 .\" redundant text (although the check is in the kernel), since,
127 .\" in another check, 'put_old' is required to be under 'new_root'.
131 is on the current root mount.
132 (This error covers the pathological case where
139 is not a mount point.
142 \fIput_old\fP is not at or underneath \fInew_root\fP.
145 The current root directory is not a mount point
146 (because of an earlier
150 The current root is on the rootfs (initial ramfs) mount; see NOTES.
153 Either the mount point at
155 or the parent mount of that mount point,
161 is a mount point and has the propagation type
165 \fInew_root\fP or \fIput_old\fP is not a directory.
168 The calling process does not have the
173 was introduced in Linux 2.3.41.
176 is Linux-specific and hence is not portable.
178 A command-line interface for this system call is provided by
182 allows the caller to switch to a new root filesystem while at the same time
183 placing the old root mount at a location under
185 from where it can subsequently be unmounted.
186 (The fact that it moves all processes that have a root directory
187 or current working directory on the old root directory to the
188 new root frees the old root directory of users,
189 allowing the old root mount to be unmounted more easily.)
193 is during system startup, when the
194 system mounts a temporary root filesystem (e.g., an
196 then mounts the real root filesystem, and eventually turns the latter into
197 the root directory of all relevant processes and threads.
198 A modern use is to set up a root filesystem during
199 the creation of a container.
203 modifies process root and current working directories in the
204 manner noted in DESCRIPTION
205 is necessary in order to prevent kernel threads from keeping the old
206 root mount busy with their root and current working directories,
207 even if they never access
208 the filesystem in any way.
210 The rootfs (initial ramfs) cannot be
212 The recommended method of changing the root filesystem in this case is
213 to delete everything in rootfs, overmount rootfs with the new root, attach
214 .IR stdin / stdout / stderr
219 Helper programs for this process exist; see
222 .SS pivot_root(\(dq.\(dq, \(dq.\(dq)
226 may be the same directory.
227 In particular, the following sequence allows a pivot-root operation
228 without needing to create and remove a temporary directory:
233 pivot_root(".", ".");
234 umount2(".", MNT_DETACH);
238 This sequence succeeds because the
240 call stacks the old root mount point
241 on top of the new root mount point at
243 At that point, the calling process's root directory and current
244 working directory refer to the new root mount point
246 During the subsequent
252 and then moves up the list of mounts stacked at
254 with the result that old root mount point is unmounted.
257 For many years, this manual page carried the following text:
261 may or may not change the current root and the current
262 working directory of any processes or threads which use the old
266 must ensure that processes with root or current working directory
267 at the old root operate correctly in either case.
268 An easy way to ensure this is to change their
269 root and current working directory to \fInew_root\fP before invoking
273 This text, written before the system call implementation was
274 even finalized in the kernel, was probably intended to warn users
275 at that time that the implementation might change before final release.
276 However, the behavior stated in DESCRIPTION
277 has remained consistent since this system call
278 was first implemented and will not change now.
281 .\" Would it be better, because simpler, to use unshare(2)
282 .\" rather than clone(2) in the example below?
283 The program below demonstrates the use of
285 inside a mount namespace that is created using
287 After pivoting to the root directory named in the program's
288 first command-line argument, the child created by
290 then executes the program named in the remaining command-line arguments.
292 We demonstrate the program by creating a directory that will serve as
293 the new root filesystem and placing a copy of the (statically linked)
295 executable in that directory.
299 $ \fBmkdir /tmp/rootfs\fP
300 $ \fBls \-id /tmp/rootfs\fP # Show inode number of new root directory
302 $ \fBcp $(which busybox) /tmp/rootfs\fP
303 $ \fBPS1=\(aqbbsh$ \(aq sudo ./pivot_root_demo /tmp/rootfs /busybox sh\fP
305 bbsh$ \fBbusybox ln busybox ln\fP
306 bbsh$ \fBln busybox echo\fP
307 bbsh$ \fBln busybox ls\fP
310 bbsh$ \fBls \-id /\fP # Compare with inode number above
312 bbsh$ \fBecho \(aqhello world\(aq\fP
320 /* pivot_root_demo.c */
327 #include <sys/wait.h>
328 #include <sys/syscall.h>
329 #include <sys/mount.h>
330 #include <sys/stat.h>
332 #include <sys/mman.h>
334 #define errExit(msg) do { perror(msg); exit(EXIT_FAILURE); \e
338 pivot_root(const char *new_root, const char *put_old)
340 return syscall(SYS_pivot_root, new_root, put_old);
343 #define STACK_SIZE (1024 * 1024)
345 static int /* Startup function for cloned child */
349 char *new_root = args[0];
350 const char *put_old = "/oldrootfs";
353 /* Ensure that \(aqnew_root\(aq and its parent mount don\(aqt have
354 shared propagation (which would cause pivot_root() to
355 return an error), and prevent propagation of mount
356 events to the initial mount namespace. */
358 if (mount(NULL, "/", NULL, MS_REC | MS_PRIVATE, NULL) == \-1)
359 errExit("mount\-MS_PRIVATE");
361 /* Ensure that \(aqnew_root\(aq is a mount point. */
363 if (mount(new_root, new_root, NULL, MS_BIND, NULL) == \-1)
364 errExit("mount\-MS_BIND");
366 /* Create directory to which old root will be pivoted. */
368 snprintf(path, sizeof(path), "%s/%s", new_root, put_old);
369 if (mkdir(path, 0777) == \-1)
372 /* And pivot the root filesystem. */
374 if (pivot_root(new_root, path) == \-1)
375 errExit("pivot_root");
377 /* Switch the current working directory to "/". */
379 if (chdir("/") == \-1)
382 /* Unmount old root and remove mount point. */
384 if (umount2(put_old, MNT_DETACH) == \-1)
386 if (rmdir(put_old) == \-1)
389 /* Execute the command specified in argv[1]... */
391 execv(args[1], &args[1]);
396 main(int argc, char *argv[])
398 /* Create a child process in a new mount namespace. */
400 char *stack = mmap(NULL, STACK_SIZE, PROT_READ | PROT_WRITE,
401 MAP_PRIVATE | MAP_ANONYMOUS | MAP_STACK, \-1, 0);
402 if (stack == MAP_FAILED)
405 if (clone(child, stack + STACK_SIZE,
406 CLONE_NEWNS | SIGCHLD, &argv[1]) == \-1)
409 /* Parent falls through to here; wait for child. */
411 if (wait(NULL) == \-1)
423 .BR mount_namespaces (7),